New Criterion for Selection in Regression Model

Abstract

The aim of this study is to propose the new criterion for selection in regression model, called NIC, and then compare the effectiveness of NIC with ten model selection criteria, namely, AIC, BIC, HQIC, AICc, AICu, HQICc, KIC, KICc_C, KICc_SB, and KICc_HM. The conditions for simulation were the differences in sample size, number of parameters in the model, regression coefficient, error variance, and distribution of independent variables. The results of the study showed that, for small to moderate sample sizes and the true model is somewhat difficult to identify, the performances of AIC and HQIC perform the best. However, they can identify the true model actually less accurate. As a result, the observed  efficiency suggests that NIC is the best criterion for small to moderate sample sizes. For the large sample size and the true model is somewhat difficult to identify, the appropriate criteria are AIC and BIC. When the sample sizes are small to moderate and the true model can be specified more easily, the appropriate criterion is NIC. For the large sample size and the true model can be specified more easily, the appropriate criterion is BIC.